Detection of 'single-leg separated' heart valves using statistical pattern recognition with the nearest neighbor classifier. Revision 1.

摘要

The goal of this work was to detect ''single-leg separated'' Bjoerk-Shiley Convexo-Concave heart valves which had been implanted in sheep. A ''single-leg separated'' heart valve contains a fracture in the outlet strut resulting in an increased risk of mechanical failure. The approach presented in this report detects such fractures by applying statistical pattern recognition with the nearest neighbor classifier to the acoustic signatures of the valve opening. This approach is discussed and results of applying it to real data are given.